Tor Research Safety Board
- What is the Tor Research Safety Board?
- What are the safety guidelines?
- How can I submit a request for advice?
- What are some examples of research that is in-scope?
- Who is on the Board?
We are a group of researchers who study Tor, and who want to minimize privacy risks while fostering a better understanding of the Tor network and its users. We aim to accomplish this goal in three ways:
- developing and maintaining a set of guidelines that researchers can use to assess the safety of their Tor research.
- giving feedback to researchers who use our guidelines to assess the safety of their planned research.
- teaching program committees about our guidelines, and encouraging reviewers to consider research safety when reviewing Tor papers.
Here's a start:
- Use a test Tor network whenever possible.
- It's safest to only attack yourself / your own traffic.
- Only collect data that is safe to make public.
- Don't collect data you don't need (minimization).
- Take reasonable security precautions, e.g. about who has access to your data sets or experimental systems.
- Limit the granularity of data (e.g. use bins or add noise).
- The benefits should outweigh the risks.
- Consider auxiliary data (e.g. third-party data sets) when assessing the risks.
- Consider whether the user meant for that data to be private.
There's plenty of room for further improvement here. In fact, we think this list itself is a really interesting research area. One of our next steps is to flesh out each of these points with a few paragraphs of explanation. Please help!
The vision is that you (the researchers) think through the safety of your planned activity, write up an assessment based on our guidelines, and send it to us. Then we look it over and advise you about how to make your plan safer, how to make your arguments crisper, or what parts really seem too dangerous to do. Later (e.g. when your paper gets published) we'll encourage you to make your assessment public. Over time we'll grow a library of success cases, which will provide best practices guidance for being safe, and also provide templates for writing good assessments.
We hope that going through this process will help you think clearly about the benefits and risks of your experiment. Hopefully our feedback on your thoughts will help too. At the same time, this process will help Tor by letting us know what research is happening — which in turn can help you, since we might be able to let you know about a concurrent experiment (with their permission of course) that will mess up your results. Also, we can let you know about upcoming Tor design changes that might influence your plans or analysis.
To best help you, we want to hear about four aspects of your proposed experiment or research plan:
- What are you trying to learn, and why is that useful for the world? That is, what are the hoped-for benefits of your experiment?
- What exactly is your plan? That is, what are the steps of your experiment, what will you collect, how will you keep it safe, and so on.
- What attacks or risks might be introduced or assisted because of your actions or your data sets, and how well do you resolve each of them? Use the "safety guidelines" above to help in the brainstorming and analysis.
- Walk us through why the benefits from item 1 outweigh the remaining risks from item 3: why is this plan worthwhile despite the remaining risks?
We encourage you to include your assessment as a section of your research paper: one of the goals here is that reviewers on program committees come to expect a section in Tor papers that explains what mechanisms the researchers used for ensuring that privacy risks were handled, and argues that the balance between new understanding and risk is worthwhile. For space reasons, you might include a streamlined version in the main body of the paper and a more detailed version in an appendix.
In the future, we'd like to come up with a more thorough template for self-assessments, to help you make sure you don't miss any critical areas. Please let us know what would help you most.
Contact address coming soon. In the meantime, mailing Roger is not a terrible plan.
This is where the templates and example self-assessments will go.
The current people who have expressed interest in the board are:
- George Danezis
- Roger Dingledine
- Tariq Elahi
- Bryan Ford
- Ian Goldberg
- Rob Jansen
- Aaron Johnson
- Damon McCoy
- Wendy Seltzer
- Micah Sherr
- Paul Syverson
Why now? The importance of Tor is growing in the world, and interest from researchers remains high as ever. Each week we run across a new paper that maybe didn't think things through in terms of keeping their users safe. We've seen lately that simply having a sensitive data set, even if you plan to never give it out, can put users at real risk. At the same time, we've seen exciting papers like PrivEx, which show that studying how to do research safely can be a research field in itself. Now is the perfect time for us to work to shape future research so we build habits of safety in our community, and so we help people to understand what is possible.
What about bad people who don't care about being safe? A safety board cannot by itself stop all dangerous Tor research. Plenty of people out there don't play the academic game, and some people don't care about user safety at all. Our goal here is to support the people who want to cooperate, while showing to the world that it's possible to do good Tor research safely.
Can't I just run Tor relays and do my experiment without telling you? Please don't! The directory authority operators have been much more conservative lately (after the CMU incident in particular) in terms of looking for suspicious patterns or behavior, and removing suspicious relays from the network. If the directory authority operators know about you, understand your research, and can read about why the benefits are worth the risks in your case, they will likely leave your relays in place, rather than surprising you by kicking your relays out of the network mid experiment.
Can I do this assessment and review process even if I'm not writing an academic paper? Please do! Our goal as stated above is "to minimize privacy risks while fostering a better understanding of the Tor network and its users". If your end goal is something other than a research paper, that's great too.
Is this an ethics board? We framed this idea as a safety board, not an ethics board. We think safety is a narrower scope: we aim to describe how to be safe, and we aim to make it the norm that reviewers and program committees expect to see an analysis of why an experiment or measurement is safe. We also are not adding new bottlenecks to the research process, such as mandating that we have to vet the analysis first — that's ultimately between the researchers and the program committees. We aren't trying to replace IRBs or other projects like ethicalresearch.org.
So I still need to go to my IRB? This safety board is orthogonal to the IRB concept. We hope that the evaluation process here will help you organize your thoughts for your IRB, but it does not replace your IRB process (if you have one).
What about confidentiality? Unless you tell us otherwise, we will keep assessments that we receive confidential in the same way that program committees do. You're coming to us much earlier in the process (ideally before the research is performed and before the paper is written), which we recognize requires more trust. We hope we add enough value to your research that you find this tradeoff worthwhile.
How will you know the right balance between benefits and risks? This is a tough one. We want Tor to get stronger long-term, but we don't want to put people into danger to do it. One answer is that most of this board is made up of professors and other online safety, security, and privacy researchers, who can provide a more neutral perspective on the right balance. The other answer is that this process is a feedback loop and we can adapt as we go: once successful assessments have gone up on this page, and people are including assessments in their papers, everybody else can look at them and decide if they used the right balance.
So you want conferences to adopt your guidelines? Not quite. We would be sad if program chairs told their reviewers "Make sure the paper follows Tor's guidelines for safe research." We would instead like the chairs to tell the reviewers "Make sure the paper has performed safe research. If you're unsure what that means, I encourage you to read Tor's guidelines to get ideas on what to consider." That is, we want the reviewers to always be thinking through, for each paper, whether this is a safe or unsafe situation. Reviewers should enforce the ethical requirements of the conference they're reviewing for — or their own ethical principles, if the conference neglected to have an opinion on the topic. Our goal here is to help them think through what to look for.
Is Tor going to do this assessment process for its own design decisions and statistics collection? Absolutely! You'll notice a big improvement over the years between our early statistics collection choices and our later ones. That learning process is part of what led to this safety board. We'd like to revisit many of Tor's design choices, especially once we've worked through some other examples here. We'd love to have your help there.